Although total worldwide foundry utilization remains below 70 percent, wafers on advanced technology (0.15 micron and below) are becoming scarce. This technology is available from a few large foundries-primarily IBM, TSMC and UMC-all of which are operating near peak capacity.

Taiwan Semiconductor Manufacturing Co., for example, is above 90 percent of total capacity and, according to its own metrics, above 100 percent capacity for 0.15 micron and below. This advanced technology comprises about 40 percent of total capacity, but that is not enough to satisfy growing interest in the reduced power and cost of 0.15- and 0.13-micron chips.

We are in this bind because, during the downturn, the large foundries underinvested in new equipment. At the time, these companies were able to meet the lower demand with existing equipment.

With demand reaching capacity limits, foundries are beginning to invest in new capacity. TSMC plans to spend $1.8 billion this year, 50 percent more than in 2003. But since it can take a year or more to bring a new fab line to production, the investment will be too late to ease the current shortages.

As a result, leading-edge capacity will be tight throughout 2004. Planned increases in capacity will be absorbed by burgeoning demand for the DSPs, microprocessors and wireless chips that power the most popular new devices.

The smaller fabless IC vendors will be the hardest hit. Unless they have long-term contracts, they will end up at the bottom of wafer allocation lists. The wafers they get will be more expensive, driving up costs.

Even the larger fabless companies will be impacted. They will see longer turnaround times and may have difficulty increasing orders to meet sudden growth in demand.

In the near term, rising prices for semiconductors will increase revenue for the chip industry. But these increases will affect the price of consumer electronics and other equipment that uses semiconductors, dampening interest in those products and slowing growth.

By next year, the investment in new equipment will finally pay off in significant increases in foundry capacity. Some experts, however, are already predicting that semiconductor growth will cool in 2005. If so, foundries will be left with excess capacity, leading to financial losses and underinvestment, and setting the stage for another boom-and-bust cycle.

Linley Gwennap is founder and principal analyst of The Linley Group (www.linleygroup.com/npu).